UMG v. Anthropic: Can International Copyright Laws Guide U.S. Law?

Sammi DietrichNews & Insights

On October 18, 2023, several music publishing companies, including Universal Music Publishing Group, Concord Music Group, and ABKCO (the “Publishers”) filed a copyright lawsuit against AI developer Anthropic in a Tennessee federal court.[1] The Publishers allege that Anthropic’s AI chatbot, “Claude,” unlawfully copied and disseminated copyrighted works, including lyrics “owned or controlled by the Publishers.”[2] This is nothing new—several companies have been sued for allegedly using unauthorized copyrighted material to train AI generative models.[3] Notably, United States copyright law is not current with the rapidly evolving pace of AI capabilities. While Congress has yet to pass any laws regarding this issue, international laws may prove persuasive in establishing a precedent for this issue.

Copyright infringement exists where there is a valid copyright and unauthorized copying.[4] Currently, only works created by humans are protected by United States copyright law.[5] The Publishers allege that Anthropic violated United States copyright law by building “its AI models by scraping and ingesting massive amounts of text from the internet . . . and then using that vast corpus to train its AI models and generate output based on this copied text.”[6]  For example, the Publishers claimed that when users prompt Claude to “‘[w]rite me a song about the death of Buddy Holly,’ the AI model responds by generating output that copies directly from the song ‘American Pie’ written by Don McLean, in violation of Universal’s copyright.”[7]

Anthropic responded to the complaint, outlining several arguments why it should not be held liable for copyright infringement. First, Anthropic argues that it has safeguards in place to prevent lyric-copying.[8] For example, Anthropic currently bars Claude from providing the full lyrics of songs, and instead provides a notice of potential infringement to the user.[9] Second, Anthropic asserts that users are actually liable for copyright infringement because they are the ones prompting Claude to generate copyrighted material.[10] Third, Anthropic argues that if it is responsible for using copyrighted work, its training falls under the “fair use” doctrine, which permits limited portions of a work for specific purposes, such as commentary, criticism, news reporting, and scholarly reports.[11] Anthropic maintains that sharing lyrics is purely for the educational purpose of helping people understand and learn from the lyrics to create better music independently.[12]

The United States has not passed any laws on the use of copyrighted materials in training AI.[13] Indeed, only ten states have enacted any legislation addressing AI thus far.[14] Some states have enacted data protection and transparency acts, requiring companies to provide consumers with the choice of opting out of relinquishing their data, and to provide disclosures whenever AI training is used.[15]

With such a limited framework to address AI concerns, the United States may look to international laws as persuasive authority on the matter.[16] For instance, India recently passed a law stating the AI Developers must get authorization to use copyrighted material to train their programs.[17] Specifically, developers need permission only if their program is going to be used for commercial purposes, aside from fair use exceptions.[18] Additionally, the European Union’s AI Act, which is heading for a vote in the European Parliament, may “become the democratic world’s first comprehensive set of laws governing the development and use of AI technologies.”[19] This law would require developers to provide a detailed summary of content used for training if the model is a “general purpose model.”[20] Although the law does not clarify the definition of a “general purpose model,”[21] it may still provide valuable legislative language for how to address the challenges of AI training models. 

AI training with copyrighted materials is still unchartered territory in United States federal law; however, international policies may be useful in guiding United States policy.  The outcome of this case may set some precedent on how courts can handle unauthorized copyright use in AI generative models.  In the meantime, everyone is keeping a watchful eye on AI trends globally.

[1] Compl., Concord Music Group, Inc. v. Anthropic PBC, Case No. 3:23-cv-01092, (M.D. Tenn.) [ECF No. 1].

[2] Id. at 3.

[3] See e.g., Emma Aistrop, The Generative AI Battle Continues: Comedian Sarah Silverman May Have Won the Fight, but Will OpenAI Win the War?, ELR News and Insights (November 1, 2023),

[4] Copyright Litigation 101, Thomson Reuters (Dec. 16, 2022),

[5] Ellen Glover, AI-Generated Content and Copyright Law: What We Know, Built In (Feb. 18, 2024),

[6] Compl., at 4.

[7] Compl., at 29-30.

[8] Daniel Tencer, Anthropic Trained its AI to Rip Off Copyrighted Lyrics, Music Publishers Allege in Escalating Court Battle, Music Business Worldwide (Feb. 15, 2024),

[9] Id.

[10] Id.

[11] Fair Use, Columbia University Libraries (Mar. 15, 2024),

[12] Claude Ai, Anthropic Fights Back Against Universal Music Group in AI Copyright Lawsuit, Claude AI (Jan. 27, 2024),

[13] Court Finds AI-Generated Work Not Copyrightable for Failure to Meet “Human Authorship” Requirement – But Questions Remain, Jones Day (Aug. 2023),

[14] U.S. State-By-State AI Legislation Snapshot, BCLP (Mar. 15, 2024),

[15] Id.

[16] Persuasive Authority, Cornell Law School (Mar. 15, 2024),

[17] Daniel Tencer, AI Developers Must Get Authorization to Use Copyrighted Materials, India’s Government Declares, Music Business Worldwide (Feb. 13, 2024),

[18] Id.

[19] Daniel Tencer, EU AI Act Headed for Parliamentary Vote, With Some Copyright Provisions Still Unclear, Music Business Worldwide (Feb. 14, 2024),

[20] Id.

[21] Id.